ERROR when training with MSRA about trihorn-net

[WJJLBJ]

As I know there should not be 000414_depth.bin in that dir

[TheoCarme]

Hello @mrezaei92 I would greatly appreciate if you could help us on that matter.
Thank you in advance.

[TheoCarme]

Actually I found the cause of the problem. It comes from the files "msra_test_list.txt" and "msra_test_groundtruth_label.txt" which respectively list the names of all the ".bin" files in the dataset and the ground truth corresponding. So I just deleted in those files the lines corresponding to the ".bin" files I didn't have in my dataset.

Also @WJJLBJ if you didn't give up at the time and resolve the issue, maybe you also succeeded in using the hand pose estimator and you could answer to my question here.

[pijiangdada]

Hello @mrezaei92 I would greatly appreciate if you could help us on that matter. Thank you in advance.

Actually I found the cause of the problem. It comes from the files "msra_test_list.txt" and "msra_test_groundtruth_label.txt" which respectively list the names of all the ".bin" files in the dataset and the ground truth corresponding. So I just deleted in those files the lines corresponding to the ".bin" files I didn't have in my dataset.

Also @WJJLBJ if you didn't give up at the time and resolve the issue, maybe you also succeeded in using the hand pose estimator and you could answer to my question here.

Can you share the "msra_test_groundtruth_label. txt" and "msra_test_list. txt" files after deleting which lines? If I delete them one by one, it would be a waste of time. If you could share them, I would greatly appreciate it!

[TheoCarme]

You will find the files here.
Just to be clear I only remove the lines from the files that did not match with an element of the dataset. I obviously did not add lines (since I did not had the ground truth) that matched with elements of the dataset and were not present.

Also did you change your profile picture just to coax me to share you the files ^^ ?

[pijiangdada]

You will find the files here. Just to be clear I only remove the lines from the files that did not match with an element of the dataset. I obviously did not add lines (since I did not had the ground truth) that matched with elements of the dataset and were not present.

Also did you change your profile picture just to coax me to share you the files ^^ ?

Thank you very much for sharing! This picture is from the anime Star Cowboy, it is an excellent work, I would like to recommend it!

After changing these two txt files, it seems that the model can run smoothly! Since only about a few dozen samples were missing, I guess there shouldn't be any impact on the training of the model. As I am a beginner in deep learning, I would like to ask again, will deleting the corresponding row cause the mismatch between the independent and dependent variables, resulting in inconsistency between the original data and labels?

[TheoCarme]

My picture is from the same anime, hence my "question".

I am not sure to understand your question, but here is my answer. Since I deleted the lines that had no match in the datasets in both files and that the "msra_test_groundtruth_label. txt" file as I understand it is containing what your calling the labels, I do not think there would be a problem.

[pijiangdada]

My picture is from the same anime, hence my "question".

I am not sure to understand your question, but here is my answer. Since I deleted the lines that had no match in the datasets in both files and that the "msra_test_groundtruth_label. txt" file as I understand it is containing what your calling the labels, I do not think there would be a problem.

Oh, I enlarged the picture and recognized it as Ed. Haha, thank you very much again for your answer!